My 2022 letter will probably be my last. Now I have to devise another use for this site. Since I’ve just spent a few days in Singapore, joining various roundtables to discuss China, I thought I would write up the notes I presented on.
Most of my remarks focus on technology. But with China, one must usually start with politics to set the scene. More than half a year has passed since the 20th Party Congress, when top leader Xi Jinping unveiled a new Politburo that overflows with his apostles. That sets up five more years of more forceful political centralization, as well as making it more than probable that Xi will be in office until 2032. Among other things, the congress would deliver the finishing blow to the formerly-thriving political entertainm…
My 2022 letter will probably be my last. Now I have to devise another use for this site. Since I’ve just spent a few days in Singapore, joining various roundtables to discuss China, I thought I would write up the notes I presented on.
Most of my remarks focus on technology. But with China, one must usually start with politics to set the scene. More than half a year has passed since the 20th Party Congress, when top leader Xi Jinping unveiled a new Politburo that overflows with his apostles. That sets up five more years of more forceful political centralization, as well as making it more than probable that Xi will be in office until 2032. Among other things, the congress would deliver the finishing blow to the formerly-thriving political entertainment business, which had spun so many beguiling and delightful tales of which elites will leap at last to oppose him.
There’s a hope that the new premier Li Qiang would make a fine truth-teller to Xi. This pro-business former party secretary of Shanghai, the thinking goes, would leverage his long relationship with Xi to change his mind. It’s hard to bury this idea. But I believe that Li did not get to where he is by vigorously speaking truth to power. In his first remarks to the media, he stated that he saw his role as a faithful implementor of the decisions of the Central Committee. 1 That does not make it sound like he has much of his own mind.
Now we contend with a Chinese leadership that has even less tolerance for elite disagreement, which probably isn’t augmenting the space for policy debates. The dangers of political centralization are easy to identify. What are the upsides? I’m finding it harder to see them. Xi has consolidated power, not apparently because he sees long-standing policy logjams he wishes to break — but for its own sake. Unlike Deng, he is not driven by an urgent desire for economic reform. Instead, he seems to view his main task as notching up the national-security consciousness of the nation. Xi achieved the impossible task of disciplining the party and silencing political opposition; there isn’t much sign that he is pursuing a policy goal commensurate with such immense political weight.
Politics didn’t need to take center stage so long as the economy could deliver its ravishing spectacles. But growth now resembles an aging star, whose act keeps being stolen by the ghastly presence of the communist state. China’s long-term economic challenges are obvious: demographic drags, a peak in property demand, debt overhangs, and a western world intent on some degree of decoupling. The surprise is that the economy hit the skids only six months after the abandonment of zero-Covid.
At the beginning of 2023, people looked forward to the economic growth that would accompany re-opening. Beijing’s overly-modest growth target of “around 5%” was going to be easy. Today, it’s surprising that the property market has better not recovered and that consumers are not spending more. Neither the service sector not the industrial sector is picking up enough steam, exacerbating a problem of youth unemployment.2 So far this year, the CSI300 is about flat. That means that China’s main stock index is at roughly the level of three other points in time: December 2022, before the economic momentum of re-opening; April 2022, when Shanghai (China’s largest city and main manufacturing hub) succumbed to full lockdown; and March 2020, when the novel coronavirus began to spread around the world. Not a lot of corporate optimism, in other words, for the outlook on profits. On the bond front, China has seen six consecutive months of outflows this year.
Can rapid growth resume? Sure. One can count up to a half-dozen times since Reform and Opening that China’s leadership decisively restarted the economic engine. There is far less latent potential that Beijing can tap into than in earlier periods (say, before its accession to the WTO), but there is still plenty of underlying demand. At the moment, however, the growth story isn’t in great shape. And one of the trends I point out in my 2022 letter is that faltering economic growth is going to feed into a domineering political agenda, and vice versa. Though it is impossible to measure, I expect that political tightening is going to constitute a meaningful drag to long-term growth.
The momentum in tech
China’s tech sector is being weighed down by slower economic growth and blows from both the US and Chinese governments. But these are not the whole story, for technology can be carried by a momentum of its own. I see China’s tech development to be, as usual, a mixed bag: some parts going poorly, other parts quite splendidly.
Start with the most thrilling new development of the past year: on generative AI, there’s not much we can use from Chinese firms. The starting point of any discussion of AI in China must be that domestic firms have failed to broadly release their reply to ChatGPT half a year since Americans have started to play with it. Yes, Chinese tech companies are developing their own generative AI tools, often scoring impressively on technical benchmarks. But they have released them in controlled settings, not to the general public.
I think the reason they haven’t given everyone access to AI chatbots is straightforward: regulators in Beijing would rather not let them run in the wild. Chinese tech companies may be hobbled by lack of access to the most advanced chips; and they’re probably hurt by the lack of training data, since most of the trainable texts are in English rather than Chinese.3 But Chinese companies rarely hesitate to release substandard products into the market in order to claim early-mover advantages, so a greater force must be holding them back.
That great force is the will of the hard men of Beijing. I suspect the Chinese leadership views large language models as something akin to social media platforms: technologies with little economic upside and significant political risk. Social media platforms like Twitter and TikTok are not increasing TFP. (Personally speaking, these platforms are horribly detracting my productivity.) Instead, they are freewheeling platforms for expression, with the potential to create political unrest. Xi and the rest of the Politburo have little reason to put AI chatbots — which only sometimes follow their guardrails — into the hands of every citizen.
At the moment, it might not be an absurd belief to treat generative AI as a toy whose value is somewhere between economically useless to socially destabilizing. But that belief may not stay reasonable for long. Americans are now integrating these tools into their lives. And they will start showing up in productivity statistics, perhaps even soon. The longer that most Chinese are unable to work with them, the greater the risk that China will be left behind in some way.
AI of course means more than chatbots and image generators. And Beijing is certainly employing a great deal of AI — but for the purposes of censorship, facial recognition, and other means for control. Rather than letting the people tinker with these technologies, the state is guarding them for itself. A question on my mind is whether this present prohibition on the broad consumer uses of AI will slow down more strategic deployments. In any case, I propose a minimum benchmark: if by the end of 2023, Chinese consumers are still mostly unable to access homegrown alternatives to ChatGPT, then at least we can revise Kai-Fu Lee’s case that while America leads on the innovation of AI, China is better positioned to lead on its implementation.
Though China is without reply to novel AI technologies, the US should stay vigilant in a protracted technological contest with a peer competitor. That is the premise of an op-ed I’ve just written for the New York Times. Summary: “If there is ever a serious disruption to trade, it’s far from obvious that American prowess in AI will overcome China’s strength of a large and adaptive manufacturing base.”
It is the weight of this large and adaptive manufacturing base that buttresses my constructive view on China’s technology development. China continues to suffer weaknesses in semiconductors, aviation, and a few other strategic technologies. But it is gaining strength in so many other sectors. My favorite examples to cite are three. Due to its prowess in electric vehicles, China is on track to surpass Japan as the world’s largest auto exporter in 2023, after edging out Germany last year. For the iPhone, China moved from a ~3% contribution to value-added in 2008 to ~25% in recent years. And where it comes to clean technologies, China has built a commanding lead. It dominates most of the supply chain in solar — from upstream polysilicon refining to downstream photovoltaic assembly — as well as much of the electric vehicle battery supply chain.
China retains considerable strengths, one which doesn’t depend so much on slowing economic growth or demographic drags. The main one is its entrenched workforce that continues to advance manufacturing complexity. I think about the humming engine that is outlined by Kevin Kelly’s concept of the “technium.” 4 That describes an ecosystem of intertwined, co-dependent, and complex technologies with a mind of its own. Though China’s leadership has grown so sour, the country’s manufacturing ecosystems continue to gain complexity, as well as global market share in many products.
Even relatively-low levels of economic growth permits considerable technological catch-up. One must always keep in mind that the Chinese task is to follow a technological ladder that western firms have laid down. They do not need to make theoretical breakthroughs to re-invent existing technologies. And neither does demographic decline guarantee a breakdown to technological momentum. Every year, nearly twice as many PhDs in STEM fields graduate from Chinese universities than American universities. It is a relatively small percentage of the population that counts for technological competition.
So I continue to hold the view that China will mostly patch up its strategic deficiencies, including in chips and aviation. That does not mean however that its firms will become the full innovative peers of the likes of ASML and TSMC. Chinese competitors are almost certainly going to be less organized and less profitable. But they will produce good enough products, lagging behind global leaders by a few years, leaving China at not too serious of a disadvantage. Chips not powerful enough to fit into the latest iPhone, perhaps, but good enough for most electric vehicles; planes not as efficient as the latest from Airbus, but good enough to fly between Shanghai and Shenzhen.
That means the US should not slacken its focus on technological competition. For America needs to figure out so much more than advanced chips. It’s on the right path after passage of the Chips Act and the Inflation Reduction Act. But they are still only starts, meaning little if the US fails to build a more rich industrial ecosystem.
A brief note on clean tech: the US is in the strange position of trying to engage in technological catch-up with a lower-wage competitor. My base case is that the IRA will mostly succeed. The amount of tax credits are so generous that America will at least be able to meet domestic demand with domestic production. (A higher threshold of success — to challenge Chinese firms in global markets — would probably be too tall an order.) Even so, there are going to be bumps on this road. One risk is that America will build lots of factories, but most are stuck at low scales of production. Apple’s Mac Pro facility in Texas, for example, didn’t really manage to scale. Another risk is that the US will lose the political will to fund these technologies for many years. Plenty of solar and battery startups are going to fail; it would be a shame if folks in Congress respond then by mocking these efforts and withdrawing the funds.
Halting technological momentum
Back to China. A gradual slowdown in economic growth won’t break technological momentum. But politics might.
Start with the external environment: fewer large markets are open to Chinese technology exports than ten years ago. For any Chinese product that might rise to the attention of a Congressperson, the US is fairly hostile. Europe remains open, but it too is grumbling about protection. A huge blow to Chinese tech firms in recent years was the loss of the Indian market. One of the many surprises 2020 was the deadly skirmish between Chinese and Indian troops that erupted after decades of relative calm. In the aftermath of the brawl, India’s government locked Chinese companies out of a market many staked growth plans on. India is not fully closed, and Chinese firms still have a lot of markets to export to. But that set has shrunk, and who can be sure that Beijing’s diplomatic and military posture won’t hurt markets for other entrepreneurs?
The west is starting to replace talk of “decoupling” with “de-risking.” I find the latter to be a marvelously Chinese word: full of ambiguity, allowing western countries to cross the river by feeling for the stones. Paul Gewirtz points out that de-risking could mean both reducing risks or eliminating them5. Which suggests, ultimately, that decoupling and de-risking could be distinctions without a difference, the latter a polite rebranding of the former.
Internal political changes are at work too. The party congress unveiled a new body, the Central Commission for Science and Technology, to be the top coordinator on tech development. The state has thus decisively shifted to a top-down approach to solve its technology problems, treating chips as the development of spaceflight or the bomb. On balance, this state-driven shift is fairly negative for overall tech development. Beijing is fundamentally misunderstanding the chip industry if it can believe that semiconductors can be run as a national space project.
Comparing which technology is “harder” is a brittle sort of game. Nonetheless, I think that chips are much more challenging than rockets. Chipmakers have to work together with both upstream suppliers and downstream customers in an R&D-intensive commercial ecosystem; the fundamental task of space agencies is to boost a heavy object skywards (without always caring where they fall down6), while controlling every aspect of the supply chain. Chips are thus commercially and technologically more sophisticated than rocket science. And it’s hard to see how it would gain from being run as the latter.
The best case for this top-down approach is that the state’s scientists can successfully identify all the crucial technologies that China must master, and that they can apply skillful policy coordination to push these developments further than they could have gone otherwise. The history of central planning does not encourage this sort of confidence. But perhaps it can work out. Even in that case, however, these scientists are likely to miss the next big thing. Beijing’s present approach to science is to lavishly fund areas critical to the state, while leaving scraps to the rest. One gap that approach missed? The mRNA vaccine. Such a technology was not really on the radar of the bureaucrats in Beijing. Instead, the mRNA is much more a validation of the American system, in which scientists play on many different fringes, prepared to scale up novel technologies during a crisis.
The general political environment poses perhaps the greatest threat to technological momentum. It’s not that autocratic regimes cannot push forward the technological frontier: late-19th century Prussia, mid-20th century USSR, and plenty of others have. But I grow less certain that a third-Xi term China will sustain an innovative drive.
A lot of entrepreneurial Chinese are unsure too. The most startling news story I read this year is that rising numbers of Chinese nationals are being apprehended at the Mexican border, trying to make the crossing into the United States7. I had not imagined that some Chinese would find such a harrowing trip to be worthwhile. That comes on top of the well-reported trend that many Chinese entrepreneurs have decamped to other markets. In the last few months, I’ve chatted with a good number of Chinese undergrads in the US, who almost to a person tell me that their parents are urging them not to return to the mainland. These groups make up a miniscule percentage of China’s population. But tech development depends on them too.
The national security attitude has prompted the creation of new laws around espionage and greater restrictiveness on data transfers. Foreigners dealing with China are feeling spooked that state secrets now encompass data on economic development, science and technology, and “other matters” designated by administrative departments8. I still wonder why Beijing has decided that now was the time to formalize these things. A foreign spy in China is hardly going to be deterred from their activities now that the state declared them illegal.
This legislation, various exit bans, and questioning of multinational businesses serve only to make the rest of the world less enthused to deal with China. Foreign investors and businesspeople are reporting that it’s harder to hear frank views even when they visit. Unless it’s with friends they’ve known for many years, they’re liable to hear the party line even from their own local employees. Sometimes, the local staff refuse even to share data with headquarters, citing domestic laws. Perhaps no wonder that investor interest in China is now a rare thing, and that outflows might surpass investment inflows this year.9
The hopeful story on Xi is that he will relax his stranglehold on society once he realizes that economic growth is important, either for his mandate or China’s prestige. The problem is that I’ve been hearing that story for years without a substantial course correction. Sometimes Xi makes tactical adjustments. But between more growth and more control, he chooses the latter nine times out of ten. For those who believe that he will wake up to the importance of growth, keep in mind that he has spent much of the past ten years talking down the pursuit of GDP.10
A fatigued people and an overbearing state aren’t best for achieving economic miracles. But they are not yet pulling the plug on technological momentum. The trouble is that Xi has given a few good tugs on that cord. And his third term is only still just getting started, with the policymaking apparatus even more intensely dependent on his whims that before.
China’s first Dongfeng missile, from a design licensed from the Soviets. We like to use “rocket science” as if it’s the hardest of the disciplines. But its sophistication is less than the semiconductor industry’s, in which companies push forward technological frontiers in a competitive commercial ecosystem. Rockets however look so much more glamorous than chips; magazine editors bemoan that they can’t do much with square circuits the size of a thumb, or technicians wearing bunny suits in oddly-lit rooms. And so we see a lot more images of rockets instead. (Photo credit from Wikipedia via Flickr)
His comments here. A friend points out, incidentally, that the premier is formally empowered by the National People’s Congress, so it’s strange for him to invoke a party body.↩ 1.
A discussion from Baiguan↩ 1.
More discussion in this Foreign Affairs piece↩ 1.
Readers should pick up Kevin’s book What Technology Wants; or for the short version of his ideas, see his recent interview with Noah↩ 1.
China apparently cares even less than most about uncontrolled re-entry of rocket debris↩ 1.
A full list of the seven categories is here at China Law Translate↩ 1.
See Andrew Batson’s discussion on how some cadres are still not getting the message↩